Our API, as well as OpenVDS's API operate primarily on floats. However, we do a lot of intermediate floating point arithmetic before we pass values along to OpenVDS. E.g. coordinate transformations. To avoid precision loss (accumulating precision loss) we do most of these calculations on double and convert back to float only when passing the values along to OpenVDS. However, this annotation to voxel transform computes on floats, and when used here it introduces a precision loss to the surface depths we are fetching from OpenVDS.
To fix this we should do this annotation to voxel transformation on doubles
Our API, as well as OpenVDS's API operate primarily on floats. However, we do a lot of intermediate floating point arithmetic before we pass values along to OpenVDS. E.g. coordinate transformations. To avoid precision loss (accumulating precision loss) we do most of these calculations on double and convert back to float only when passing the values along to OpenVDS. However, this annotation to voxel transform computes on floats, and when used here it introduces a precision loss to the surface depths we are fetching from OpenVDS.
To fix this we should do this annotation to voxel transformation on doubles